Abstract 5704: An integrated immune signature predictive of adjuvant immunotherapeutic benefits for high-risk melanoma

Abstract

Abstract Introduction CTLA4 blockade with ipilimumab was more favorable than interferon-α2b (IFN) in high-risk melanoma in phase adjuvant III trial E1609. Characterization of the pretreatment tumor immune biomarkers and clinical covariates may inform the likelihood of response to ipilimumab and other immune checkpoint inhibitors (ICI), and guide future development of this and other modalities in this patient population. Methods We utilized PATH-SURVEIOR, a bioinformatics framework developed in-house for associating genes and pathway signatures with clinical endpoints, to perform survival analysis of gene expression levels of 31 candidate immune-related biomarkers based on previous preliminary data. We analyzed microarray gene expression data from 471 melanoma patients treated with ipilimumab (ipi) and 248 melanoma patients treated with IFN as part of E1609. We then developed a LASSO Cox regression model and validated our model in 22 patients treated with neoadjuvant ipi in a separate clinical trial. Results Using PATH-SURVEIOR, we evaluated 31 candidate immune biomarkers and their association with patient outcome by including treatment group (ipi and IFN) as a multiplicative covariate interaction in the Cox hazard model. Our analysis identified CXCL9, CD8A, CXCL10, and INPP5D as Tier 1 biomarkers (HR > 1 and P < 0.05) and IDO1, IGKC, and IL2RB as Tier 2 biomarkers (HR > 1 and P < 0.1). Next, we developed an ipilimumab immune-based risk score using LASSO Cox regression (L-IPI7) based on these 7 aggregate biomarkers. We then split our 471 ipi-treated cohort into training (310, 66%) and testing (161, 33%) cohorts and assessed our model for its ability to predict overall survival (OS) and relapse-free survival (RFS). Our risk score was capable of stratifying ipi-treated patients into High-Risk and Low-Risk populations, which correlated with OS. As a negative control, we assessed our risk score in 248 IFN-treated patients and found no significant association with OS. As validation, we applied our L-IPI7 score to a cohort of 22 patients treated with neoadjuvant ipi and determined that the score was able to predict patients with a high risk of relapse. Interestingly, when we developed an interactive Cox-regression model with colitis status (grade 0-1 vs grade 2+), we found that neoadjuvant ipi patients with low-grade colitis were associated with a higher L-IPI7 risk score for disease relapse. In addition, we determined that: i) higher age and higher L-IPI7 risk score identified patients with the worst OS and RFS And ii) female patients with a low L-IPI7 risk scores had a better OS and RFS. Conclusions We developed a broadly applicable model based on LASSO Cox Regression predictive of adjuvant ipi treatment outcomes in melanoma. Our L-IPI7 score based on expression of CXCL9, CD8A, CXCL10, INPP5D, IDO1, IGKC, IL2RB effectively predicts survival, with interactions with age, gender and on-treatment development of colitis. Citation Format: Alyssa N. Obermayer, Timothy I. Shaw, Sandra J. Lee, F. Stephen Hodi, William A. LaFramboise, Walter Storkus, Arivarasan D. Karunamurthy, Patrick Hwu, Howard Streicher, Dung-Tsa Chen, John M. Kirkwood, Ahmad A. Tarhini. An integrated immune signature predictive of adjuvant immunotherapeutic benefits for high-risk melanoma. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5704.